Skip to main content
. 2014 Nov 13;9(11):e112774. doi: 10.1371/journal.pone.0112774

Table 6. Top negation context features in a multi-corpus model, by chi-square value; and feature rank in domain-specific models.

Feature Rank in Training Data
Feature Description Chi∧2 all i2b2 mipacq negexts sharp
(D) DepNeg path: dt_nmod_mod 16713.1 1 5 1 2 1
(A) Bag of 5 preceding words: no 15601.1 2 1 3 3 2
(E) Tree Fragment Context Above-Left: (DT no) 15263.2 3 3 2 4 3
(A) Bag of 10 preceding words: no 14928.9 4 2 4 1 5
(A) Bag of 3 preceding words: no 14207.4 5 4 5 5 4
(B) Preceding word #0: no 10683.5 6 6 9 6 10
(C) Cue category: no 9848.3 7 7 6 12 6
(C) Cue word: no 8866.9 8 9 7 7 15
(C) Cue phrase (any negation) 8110.7 9 10 8 8 9
(E) Tree Fragment Context Above-Left:(NP (DT no) (CONCEPT )) 8038.3 10 8 18 13 23
(D) DepNeg path: negverb->dobj_mod 3817.3 11 12 13 16 285
(E) Tree Fragment Context Above-Left:(VBZ semclass_deny) 3809.4 12 13 10 15 851
(A) Bag of 10 preceding words: denies 3081.2 13 22 12 21 1195
(E) Tree Fragment Context Above-Left: (DT any) 2721.2 14 43 11 24 486
(B) Preceding word #2: no 2672.9 15 15 28 22 53
(C) Cue category: deny 2479.0 16 16 19 38 327
(A) Bag of 5 preceding words: denies 2380.3 17 28 16 26 2070
(A) Bag of 5 following words: or 2350.9 18 25 30 9 46
(E) Tree Fragment Context Above-Left: (NP (DT no) (NML )) 2247.9 19 27 44 34 19
(A) Bag of 10 following words: or 2242.1 20 26 29 10 39

Feature types are classified as in Section 3.4.